Journal of Human Evolution 130 (2019) 126e140

Contents lists available at ScienceDirect

Journal of Human Evolution

journal homepage: www.elsevier.com/locate/jhevol

A new interpretation of Madagascar's megafaunal decline: The “Subsistence Shift Hypothesis”

Laurie R. Godfrey a, *, Nick Scroxton b, c, Brooke E. Crowley d, Stephen J. Burns b, Michael R. Sutherland e, Ventura R. Perez a, Peterson Faina f, David McGee c, Lovasoa Ranivoharimanana f a Department of Anthropology, University of Massachusetts, Amherst, MA 01003, USA b Department of Geosciences, University of Massachusetts, Amherst, MA 01003, USA c Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA d Departments of Geology and Anthropology, University of Cincinnati, Cincinnati, OH 45221, USA e Department of Mathematics and Statistics, University of Massachusetts, Amherst, MA 01003, USA f Departement Bassins Sedimentaires Evolution Conservation (BEC), Universite D'Antananarivo, Antananarivo 101, Madagascar article info abstract

Article history: Fundamental disagreements remain regarding the relative importance of climate change and human Received 7 September 2018 activities as triggers for Madagascar's Holocene megafaunal extinction. We use stable isotope data from Accepted 4 March 2019 stalagmites from northwest Madagascar coupled with radiocarbon and butchery records from subfossil bones across the island to investigate relationships between megafaunal decline, climate change, and habitat modification. Archaeological and genetic evidence support human presence by 2000 years Before Keywords: Common Era (BCE). Megafaunal decline was at first slow; it hastened at ~700 Common Era (CE) and Quaternary extinction peaked between 750 and 850 CE, just before a dramatic vegetation transformation in the northwest that Paleoclimate Speleothems resulted in the replacement of C3 woodland habitat with C4 grasslands, during a period of heightened Population expansion monsoonal activity. Cut and chop marks on subfossil lemur bones reveal a shift in primary hunting Indian Ocean trade network targets from larger, now-extinct species prior to ~900 CE, to smaller, still-extant species afterwards. By 1050 CE, megafaunal populations had essentially collapsed. Neither the rapid megafaunal decline beginning ~700 CE, nor the dramatic vegetation transformation in the northwest beginning ~890 CE, was influenced by aridification. However, both roughly coincide with a major transition in human subsistence on the island from hunting/foraging to herding/farming. We offer a new hypothesis, which we call the “Subsistence Shift Hypothesis,” to explain megafaunal decline and extinction in Madagascar. This hypothesis acknowledges the importance of wild-animal hunting by early hunter/foragers, but more critically highlights negative impacts of the shift from hunting/foraging to herding/farming, settlement by new immigrant groups, and the concomitant expansion of the island's human population. The interval between 700 and 900 CE, when the pace of megafaunal decline quickened and peaked, coincided with this economic transition. While early megafaunal decline through hunting may have helped to trigger the transition, there is strong evidence that the economic shift itself hastened the crash of megafaunal populations. © 2019 Elsevier Ltd. All rights reserved.

1. Introduction habitat modification) as triggers for megafaunal extinction con- tinues to be debated (see Battistini, 1965; Dewar, 1984; Burney Whereas it is well known that Madagascar's megafauna et al., 2003, 2004; Crowley, 2010; Douglass and Zinke, 2015; declined and vanished after humans arrived, the relative impor- Burns et al., 2016; Ekblom et al., 2016; Crowley et al., 2017; tance of climate and human impacts (whether through hunting or Salmona et al., 2017; Anderson et al., 2018; Douglass et al., 2018; Hixon et al., 2018). Explanations generally focus on climatic or anthropogenic pressures, or both. Disentangling the effects of * Corresponding author. fi E-mail address: [email protected] (L.R. Godfrey). “natural” and “human-induced” change can be dif cult, especially https://doi.org/10.1016/j.jhevol.2019.03.002 0047-2484/© 2019 Elsevier Ltd. All rights reserved. L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 127 since natural aridification and human disturbance (e.g., fire, eight of which we radiocarbon dated (Crowley et al., 2017). We also deforestation, or agriculture) can have similar effects on the checked previously-dated humeri and femora of extinct lemurs in distributions and abundance of endemic plants and animals. Most the collection at the Universite d'Antananarivo for evidence of explanations for the extinction of Madagascar's megafauna posit human butchery. complex sequences of events. The data, however, do not support Our third important new source of data is calcium carbonate the proposed sequences and over the past two decades no cave deposits, or speleothems (such as stalagmites). In recent years, consensus has emerged. there has been an increased interest in investigating Madagascar's The “synergy” hypothesis (Burney et al., 2003, 2004) maintains past climate and habitat transformations using speleothems (Brook that humans were the primary triggers for extinction, although et al.,1999; Burns et al., 2016; Scroxton et al., 2017; Voarintsoa et al., periods of aridification prior to human arrival may have increased 2017). Speleothems provide an underutilized (in Madagascar) high the vulnerability of the megafauna. According to this hypothesis, resolution and precisely dated record of past climate and envi- human arrival around 2400 years ago was followed in rapid suc- ronmental changes (Fairchild and Baker, 2012). Depending on the cession by: (1) megafaunal population crash (due to megafaunal sampling resolution and degree of hydrological averaging in the hunting primarily in more open habitats), (2) the expansion of open karst host rock above the cave, speleothem records can range from habitats (due to a surge in the intensity of natural fires in grassy decadal to subannual in resolution. Recent improvements in mass biomes now uncontrolled by grazers), (3) shrinking forests (due to spectrometry techniques allow for radiometric UeTh disequilib- the impact of fire), and (4) concomitant extinction of forest- rium dating of speleothems with errors frequently less than 1%, dependent large-bodied lemurs and other creatures. The main enabling the construction of very precise chronologies (Hellstrom, inferences of this hypothesis are that the megafaunal crash 2003; Cheng et al., 2013). For example, half of the UeTh dates from occurred largely in the first part of the first millennium CE (within a stalagmite from Anjohibe Cave in northwestern Madagascar have several centuries of human arrival), and that it preceded and in 2s errors that are less than a decade (Scroxton et al., 2017). effect triggered habitat change through fire. Furthermore, it pre- In addition to providing precise, accurate chronologies, speleo- ceded the introduction of many domesticated plants and animals, thems provide continual records of changes in past vegetation and which in turn occurred with the expansion of the trade network rainfall through the use of stable carbon and oxygen isotope ratios and early spread of settlement sites almost 1000 years later. respectively (McDermott, 2004; Lachniet, 2009). Carbon isotope In contrast, Virah-Sawmy et al. (2009, 2010) argued that habitat values recorded in stalagmites can be used to identify major shifts change occurred in synchrony across Madagascar, but not because in vegetation on land overlying caves because, in karst systems, of fire set by humans; rather, it was island-wide drought that isotope signals are transferred from plants to soil waters to spe- triggered both habitat change and the megafaunal crash. According leothem carbonate (reviewed in McDermott, 2004). The d13C values to this hypothesis, the megafauna declined long after humans in stalagmites may be affected by a variety of factors, including the arrived, in association with the peak of a devastating drought. Early amount of isotope discrimination by plants, soil respiration rates, episodes of aridification in the northwest (Wang and Brook, 2013), the degree of air-solution CO2 exchange in the soil and epikarst central highlands (Gasse and Van Campo, 1998), southwest (Mahe (Genty et al., 2001; Wong and Breecker, 2015), and the extent of CO2 and Sourdat, 1972; Burney, 1993; Goodman and Rakotozafy, 1997) degassing at the stalagmite surface. Major changes in the d13C and southeast (Virah-Sawmy et al., 2009, 2010) were not always values of speleothems, however, may also reflect shifts in vegeta- synchronous. However, gradual drying between 50 and 1000 CE tion type (i.e., shifts from forest to grassland), which may or may (1900 and 950 calendar years before present, or cal BP) culminated not be climate-driven (Dorale et al., 1998; Burns et al., 2016). When in an island-wide drought between 950 and 1050 CE that was vegetation is dominated by C3 plants, speleothem carbonates can sufficiently strong to trigger megafaunal extinction (Virah-Sawmy be expected to yield d13C values between 14‰ and 6‰. In À À et al., 2009, 2010). This hypothesis posits that habitat change trig- contrast, carbonates deposited in equilibrium with CO2 respired gered megafaunal extinction, and not the opposite, as is maintained from C plants will generally exhibit d13C values between 6‰ 4 À by the synergy hypothesis. It also places the megafaunal crash and 2‰ (McDermott, 2004). Oxygen isotope values in speleo- þ approximately 1000 years later than does the synergy hypothesis. thems are also used as records of rainfall variability. In the tropics, Clearly, to evaluate these hypotheses, we must better understand and under monsoonal rainfall regimes such as those that typify the timing of the megafaunal crash, the climate of Madagascar Madagascar, speleothem d18O values typically act as a proxy for the before and during the crash, and what humans were doing at the strength of monsoon circulation: oxygen isotopes are fractionated time. Did the crash occur prior to, just after, or well after human during repeated convection and re-evaporation in the sub-cloud arrival? layer (Risi et al., 2008). As a result, under highly convective Here, we revisit the question of megafaunal extinction by monsoonal systems, the d18O values of precipitation are correlated updating and comparing three sources of data that bear on the with regional rainfall amount (Risi et al., 2008; Kurita et al., 2009; question. First, the radiocarbon record for subfossil vertebrates has LeGrande and Schmidt, 2009). Minor modification of the d18O improved significantly since the publication of Burney et al.'s signal can occur during transit through the karst and during spe- (2004) chronology for late prehistoric Madagascar (97 dates), and leothem formation, particularly in caves with low humidity and indeed since the publication of Crowley's (2010) updated chro- high airflow, which promote evaporation rather than degassing on nology (320 dates). Four hundred and thirty-eight dates are now the speleothem surface. The most reliable speleothem records are available (see Crowley and Godfrey, 2013; Crowley and Samonds, taken from caves with low airflow and high humidity. 2013; Crowley et al., 2017), 195 of which fall within the period Burns et al. (2016) produced a sub-decadal stable isotope record critical for testing extinction hypotheses e i.e., the period during from two stalagmites (AB2 and AB3) from Anjohibe cave in which time large-bodied vertebrates declined and ultimately van- northwestern Madagascar. From these stalagmites, a reproducible ished (between 50 BCE and 1350 CE). record of d13C values served as a proxy for vegetation changes Second, we offer new data on human butchery of extinct lemurs above the cave, while associated d18O records served as a proxy for during the past 2000 years. To improve on prior records, we eval- variation in rainfall. Voarintsoa et al. (2017) analyzed three addi- uated 183 humeri and femora of Pachylemur insignis from Tsirave in tional stalagmites from the same cave (MA2, MA3 and ANJ94-5), the collection of the Universite d'Antananarivo. We identified effectively corroborating many of the conclusions drawn by Burns perimortem cut and/or chop marks on several dozen individuals, et al. (2016). Using higher resolution stable oxygen isotope data 128 L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 and detailed mineralogical assessment of an intensively sampled 2.2. Records of human butchery stalagmite, AB2, Scroxton et al. (2017) have since documented changes in rainfall on a quasi-annual scale and identified climate We investigated human exploitation of extinct versus extant fluctuations at centennial scales. This was a substantial improve- terrestrial vertebrates, looking for evidence of a temporal shift. ment over the sub-decadal record produced by previous studies. Published records of dated butchered bones from Madagascar are The effects of climate and human activities can be decoupled meagre, but recent surveys and a newly-expanded radiocarbon when carbon and oxygen isotope proxies from stalagmites are database have accorded us the opportunity to add significantly to considered together. Simultaneous and coherent increases or them. Cut and chop marks on fresh bone have features that decreases in the two proxies indicate a likely rainfall control on distinguish them from scratches resulting from natural abrasion, changes in vegetation, soil and epikarst processes. In contrast, the teeth or claws of predators, or excavation trowels and shovels abrupt changes in carbon isotope values, particularly positive (Perez et al., 2005; Meador et al., 2017)(Fig. 2). We followed the excursions (signaling a change from C3- to C4-dominated habitats) methods of Perez et al. (2005) in identifying cut and chop marks. with no corresponding shifts in oxygen, indicate a non-climatic Our present analysis of butchery includes only dated bones that forcing on vegetation. exhibit the pattern of purpose associated with human processing of In this paper, we generate a new chronology for Late Holocene meat for consumption, and that fall within our search interval. Our Madagascar by combining data from our updated radiocarbon, tally therefore excludes undated butchered bones of extinct species butchery, and stalagmite chronological and isotope records. Ulti- (e.g., a chopped femur of Hadropithecus illustrated by Godfrey et al., mately, we seek to use our chronology to test hypotheses and gain 2006a, and modified Aepyornis hildebrandti from Antsirabe docu- new insights on the events that triggered the decline and disap- mented by Hansford et al., 2018), and bones exhibiting postmortem pearance of the island's megafauna. Specifically, we address the human modification, such as an elephant bird tibiotarsus from following questions: (1) What does the radiocarbon record tell us Itampolo (Burney et al., 2004). We excluded elephant bird eggshell about the trajectory and timing of Madagascar's megafaunal that is currently being tested for perimortem versus postmortem decline? (2) What does the subfossil record tell us about the tem- exploitation, and modified marine shell dated at nearly 3000 years poral context of human consumption of large versus small-bodied old, all from the Velondriake Marine Protected Area in southwest endemic species? (3) What do d18O values from stalagmites at Madagascar (Douglass, 2016). We also excluded butchered bones of Anjohibe tell us about wet and dry periods in northwestern extinct and extant species from Beanka (near Bemaraha in north- Madagascar over the period during which the megafauna crashed, western Madagascar) that are currently being analyzed and dated and relative to local changes in vegetation? (4) Can we make climate (LRG and VRP, pers. obs.). Finally, butchered specimens falling well generalizations for other regions of Madagascar on the basis of outside our search interval such as limb bones of Aepyornis max- conditions recorded in the northwest? imus at Christmas River in south central Madagascar and Muller- ornis sp. from Lamboharana on the southwest coast (Hansford et al., 2. Materials and methods 2018) were excluded from our analysis. Claims of butchery for Hippopotamus bones from Anjohibe and Palaeopropithecus bones 2.1. Trajectory of megafaunal decline from Taolambiby's Methuen collection have been challenged (cf. Perez et al., 2005; Gommery et al., 2011; Goodman and Jungers, We searched for evidence of a temporal shift in the relative 2014; Anderson et al., 2018). These bones are also outside our prevalence of extinct versus extant animals in subfossil deposits. search interval and are not further considered here. We restricted our search to the period between 50 BCE and 1350 CE We did include butchered hippopotamus and elephant bird (i.e., 2000 to 600 cal BP) as prior research demonstrates unequiv- bones from Lamboharana and Ambolisatra (MacPhee and Burney, ocally that megafaunal decline occurred during this period (Burney 1991; Hansford et al., 2018) and an extinct lemur from Manombo et al., 2004; Crowley, 2010). Furthermore, for taphonomic reasons, Toliara that fall within our search interval. Two paleontological there are very few radiocarbon-dated extant animal bones that are sites, Taolambiby and Tsirave, both in south central Madagascar, older than this (Fig. 1). Subfossil sampling is uneven across cen- have sufficient numbers of radiocarbon-dated butchered bones turies within this search interval, so tracking simple changes in within our search interval to yield information regarding “clusters” megafaunal frequency per unit time is not the best metric for un- of dates. These sites were initially excavated in the early 1900s; Paul derstanding megafaunal decline. Methuen collected at Taolambiby in 1911, and Charles Lamberton at Instead, we measured the difference between the ratios of bones Tsirave in 1930. In 1966, Alan Walker revisited Taolambiby with of extinct to extant species before and after selected times (or cut- Paul Martin. Both recognized it as a butchery site; Paul Martin points) within our full search interval (50 BCE to 1350 CE). We found pottery sherds and a carapace of a tortoise that he believed constructed a series of two-by-two tables of frequencies of extinct was artificially perforated (Raison and Verin, 1967). Walker versus extant animals every 100 years across the entire 1400-year collected bones of lemurs and other animals, some of which were period (thus, for example, before and after 50 CE, before and after butchered. Most of the subfossil bones in this collection belonged to 150 CE, before and after 250 CE, etc.), and considered the preva- extant animals; however, Walker's collection also included some lence of extinct and extant animals before and after each. We then extinct lemurs (P. insignis, Archaeolemur majori, and Palae- measured the “odds ratio” for each table by calculating the ratios of opropithecus ingens). This collection was later donated to the Uni- bones of extinct to extant species occurring before and after each versity of Massachusetts, Amherst, after which some butchered selected cut-point, and comparing the two ratios. Changes in the specimens were radiocarbon dated. A chopped distal ulna of odds ratios across the 2 by 2 tables reflect changes in the pace of Palaeopropithecus reported by Perez et al. (2005) failed to yield a megafaunal decline. They can be quantified using any of a series of date, but 12 butchered specimens of the extant lemur, Propithecus, well-known and intimately-related statistics, such as the odds did; one was modern. ratio, the odds themselves, the cross product ratio, or either of the In contrast to Taolambiby, Tsirave was not identified as a common c2 statistics (Pearson or maximum likelihood). Here, we butchery site until 2005, when Perez et al. (2005) figured a used maximum likelihood c2 statistics as our descriptive tool to butchered femur of P. insignis from this site. No artifacts have been characterize the trajectory of megafaunal decline. reported from this site. L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 129

Figure 1. Published radiocarbon records for extinct and extant taxa in subfossil deposits over the past three thousand years (modified from Crowley, 2010). White circles represent extant species. Gray circles represent extinct species. Horizontal dashed lines show the boundaries of our search interval. When both extinct and extant species belonging to a single genus have both been dated (e.g., Cryptoprocta spelea and C. ferox), single columns can include both white and gray circles. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article).

To characterize the distribution of radiocarbon dates for speci- 2.3. Using stalagmites to assess climate shifts in time and space mens within our search interval from these sites, we identified individual clusters as groups of individuals whose 14C dates fall We used five published stalagmite paleoclimate records from within ±1 standard deviation of one another. For these calculations, Anjohibe (AB2, AB3, MA3, MA2, and ANJ94-5) to assess climatic and we used the standard deviations associated with raw 14C dates, vegetative changes in northwestern Madagascar over the past two which are appropriate measures of the associated dispersion, and millennia. Three of these stalagmites (AB2, AB3, and MA3), which often considerably smaller than standard deviations associated were previously analyzed for changes in vegetation, document a with calibrated ages. shift from predominantly C3 to predominantly C4 vegetation

Figure 2. a) UA 3059 left femur of P. insignis from Tsirave, originally published by Perez et al. (2005), showing one chop mark and three cut marks (one just proximal to the chop mark and two oriented horizontally across both femoral condyles); b) UA 1451 left femur of A. majori from Manombo Toliara, with cut marks indicative of dismemberment, including one on the posteromedial portion of the neck, just distal to the femoral head, and two, also on the posteromedial aspect of the bone, proximal to the lesser trochanter; c) UA 3037 left femur of P. insignis from Tsirave, showing anterior aspect, with two chop marks on the femoral head and several cut marks on the anterior shaft, including one just distal to the lesser trochanter. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article). 130 L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 beginning in the late 800s CE. This shift was first identified in the populations were likely very small, and their further decline did d13C records of stalagmites AB2 and AB3 (Burns et al., 2016), and little to alter the extinct to extant odds in subfossil deposits. replicated in stalagmite MA3 (Voarintsoa et al., 2017). All five In summary, the period of greatest change in odds ratios began stalagmites have been analyzed for d18O values, but only one (AB2) between 650 and 750 CE (prior to which the odds decisively at exceptionally high resolution (Scroxton et al., 2017). Its d18O re- favored extinct animals) and ended between 850 and 950 CE cord was interpreted as evidence of variation in monsoonal rainfall (by which time the opposite was true). By 1050 CE, megafauna were in northwestern Madagascar over the past 1700 years (the age rare across the island. range covered by AB2; Scroxton et al., 2017). These authors used Bayesian Change Point Analysis (BCPA), a statistical treatment that 3.2. Records of human butchery or modification of fresh bone determines points in time at which changes in the mean state of isotopic variability occur (Ruggieri, 2013), to identify a series of Figure 3 shows the geographic distribution of subfossil speci- relatively wet and dry periods. mens with cut- or chop marks that fall within our search interval To determine to what extent the stalagmite records of Anjohibe (i.e., 50 BCE to 1350 CE). Dated, butchered individuals belonging to can be considered to reflect regional (as opposed to local) climate five extinct species (Hippopotamus lemerlei, A. maximus, Mullerornis variability, Scroxton et al. (2017) compared the new high-resolution sp., A. majori, and P. insignis) come from four sites (i.e., Ambolisatra, Anjohibe stalagmite data with paleoprecipitation records from two Lamboharana, Manombo-Toliara, and Tsirave) (Table 2). Butchered sites in East Africa: Lake Challa on the border of Kenya and Tanzania bones of the extant species, Propithecus verreauxi, are all from (Buckles et al., 2016), and Lake Naivasha in Kenya (Verschuren et al., Taolambiby. Of the 24 specimens listed on Table 2, 22 fall within a 2000). They found regional synchronicity in rainfall variability over narrow period from 770 to 1140 CE. the past two millennia on a multi-centennial scale. New records of butchery include an A. majori from Manombo- To assess intra-Madagascar coherency in modern rainfall (i.e., Toliara (southwestern Madagascar) and eight specimens of P. whether the climate of the northwest is representative of the island insignis from Tsirave, all of which fall within our search interval as a whole), we correlated rainfall anomalies in the region of Anjo- (Table 2). Tsirave proves to have been a site with heavy butchery hibe with rainfall anomalies across Madagascar using the CRU TS3.23 activity, and our new radiocarbon dates enable us to combine dataset (Harris et al., 2014), which interpolates climate anomalies butchery records from Tsirave with those previously collected at from over 4000 high quality weather stations and an existing Taolambiby (Perez et al., 2005) in analyzing temporal changes in (1961e1990) climate baseline to produce 0.5 gridded monthly human butchery practices. Using our single standard deviation climate data from 1901 to the present. We averaged monthly pre- cluster-search rule, seven of the eight radiocarbon-dated butchered cipitation anomalies to the hydrological year (JulyeJune) and corre- bones of Pachylemur from Tsirave form a cluster with a mean age of lated them with anomalies at the Anjohibe grid cell using the 834.3 CE and range of 825e870 CE. One individual, at 940 CE, falls Koninklijk Nederlands Meteorologisch Instituut's (KNMI) climate outside this cluster. The full date range for 41 P. insignis from Tsirave explorer (Trouet and Van Oldenborgh, 2013). Given the paucity of covers 2990 years; the clustering we observe in the modified bones historical weather station data in Madagascar, variability in the is not random. It appears that Charles Lamberton dug through a dataset is likely to be of low spatial resolution. However, we observed human occupation layer at this paleontological site without no significant change in the spatial extent of the correlation when recognizing it as such. Eleven radiocarbon-dated butchered bones using different time periods varying in length, age, or both. We of Propithecus from Taolambiby form three clusters: the first at therefore believe that the spatial correlation is not dependent on the 942.5 CE (n 6, range 930e955 CE), the second at 1023.3 CE (n 3, ¼ ¼ number of weather stations. range 1000e1065 CE) and the third at 1135 CE (n 2, range ¼ 1130e1140 CE) (Table 2). 3. Results The observed date clusters for butchered extinct and extant le- murs from Taolambiby and Tsirave, each spanning less than 100 3.1. Trajectory of megafaunal decline years, exist within sites that individually span ~4000 years (1805 BCE to modern for Taolambiby, N 71; and 1850 BCE to 1810 CE for ¼ Our database of 195 radiocarbon-dated subfossil vertebrates Tsirave, N 61). Thus, the existence of time-constrained clusters at ¼ falling between 50 BCE and 1350 CE reveals the existence of a these sites is not an artifact of limited time sampling at single sites. constrained period of exceptionally rapid change in the relative Rather, the temporal clustering of butchered specimens at Tao- frequencies of extinct to extant animals. This can be seen in the lambiby and Tsirave provides possible evidence for the existence of maximum likelihood c2 values for a sequence of two-by-two human occupation layers that were destroyed by excavation. The subtables, each comparing the frequencies of extinct and extant single cluster of butchered Pachylemur from Tsirave (825e870 CE) is subfossils before and after a single cut-point (Table 1). Each older than all three clusters of extant Propithecus from Taolambiby sub-table necessarily totals 195 individuals, as each considers all (i.e., 930e955, 1000e1065, and 1130e1140 CE). Moreover, clusters individuals within the entire 1400-year interval; they differ only in for extinct and extant lemurs fall on opposite sides of our estimated the cut-point selected for quantifying “before” and “after” odds. inversion point for extinct-extant subfossil odds (850 CE). The maximum likelihood c2 statistic increases gently (from 14.0 to 16.2 to 19.0) at cut-points 450, 550, and 650 CE, after which it 3.3. Stalagmite records of shifts in climate in space and time rises sharply (to 30.8) in one century, and then peaks at 32.7 at cut- point 850 CE (Table 1). This peak marks the transition from extinct- Figure 4a-i document the concurrent climate and vegetation dominated odds (>2.5:1, extinct to extant, prior to 850 CE) to histories reflected in stalagmites AB3 (Fig. 4a-c), AB2 (Fig. 4d-f), and extant-dominated odds (<1:1, extinct to extant, after 850 CE). The MA3 (Fig. 4g-i) at Anjohibe. A very large monotonic shift observed maximum likelihood c2 statistic then falls sharply over the next in the d13C values of stalagmites AB3 and AB2 by Burns et al. (2016) two centuries (Table 1). The steep decline in c2 values between 850 and replicated in MA3 by Voarintsoa et al. (2017) beginning in the and 950 CE (from 32.7 to 20.9) mirrors the steep rise (from 19.0 to late 800s CE was interpreted as evidence for a vegetation shift from 30.8) between 650 and 750 CE; the subsequent fall (from 20.9 to 6.9 a landscape dominated by C3 vegetation to one dominated by C4 between 950 and 1050 CE) exceeds that of the rise (from 16.2 to vegetation (Fig. 4b, e, and h, yellow box). Conversely, the d18O re- 19.0) between 550 and 650 CE. After 1050 CE, megafaunal cords from Anjohibe stalagmites (Fig. 4c, f, and i) demonstrate that L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 131

Table 1 The pace of change in extinct-to-extant subfossil odds, quantified using the maximum likelihood c2 statistic for eight 2 2 tables, each with a different temporal cut-point, Â within our search interval (50 BCE to 1350 CE).

Cutpoint (yrs CE) Time bin Frequency extinct Frequency extant Maximum likelihood c2 1150 1150 CEd1350 CE 2 4 4.334 50 BCEd1150 CE 141 48 1050 1050 CEd1350 CE 7 9 6.858 50 BCEd1050 CE 136 43 950 950 CEd1350 CE 10 18 20.884 50 BCEd950 CE 133 34 850 850 CEd1350 CE 20 29 32.718 (peak) 50 BCEd850 CE 123 23 750 750 CEd1350 CE 47 40 30.775 50 BCEd750 CE 96 12 650 650 CEd1350 CE 67 42 19.022 50 BCEd 650 CE 76 10 550 550 CEd1350 CE 78 44 16.175 50 BCEd550 CE 65 8 450 450 CEd1350 CE 93 47 13.974 50 BCEd450 CE 50 5

there was no dramatic, sustained change in rainfall at any time over Other stalagmites from Anjohibe do not show comparable d18O the past two millennia, but that some modest variability did occur increases. However, they do present other evidence for decreased at the centennial scale (Scroxton et al., 2017). precipitation such as hiatuses or decreases in growth rates. The Before the carbon excursion at 890 CE, AB3 exhibits a d13C value lack of isotopic replication suggests that any precipitation decrease around 9‰ and the new AB2 record registers similar values during the 10th and 11th centuries was modest, certainly not À at 7‰. These values are within the expected range for a landscape enough to have had major impacts on regional flora and fauna. In À with primarily C3 plant cover, such as the dry deciduous forests that any case, the decrease in precipitation observed does not reflect a occur today in pockets in the modern northwest (Burgess et al., dry period of any severity. Speleothem d18O values in stalagmite 2004). One hundred years after the beginning of the carbon AB2 during the 11th and 12th centuries are not unusually high excursion, AB3 registers a d13C value of 0‰, which is well within (dry) in comparison to other dry periods such as the 5th or the expectations for a C4 dominated landscape such as open grassland. 20th centuries (Scroxton et al., 2017). We conclude that north- The d13C record for AB2 matches the excursion exhibited over 100 western Madagascar likely underwent a period of decreased pre- years in AB3 but shows a continual rise to 3‰ or higher (indeed, cipitation in the late 10th/early 11th centuries, linking the wettest þ to 6‰) over a somewhat longer period. The magnitude of the period of the last 1700 years with a period that was drier but not þ carbon excursion, inferred from two stalagmites, was at least 8.5‰ dramatically so. in a 100-year period, and it increased to 11.5‰ or higher with time. There is a difference in timing between different stalagmite Terminal values of 2to 6‰ are highly enriched and leave no records regarding both the d13C and d18O changes: stalagmite AB2 þ þ doubt that the dominant terminal landscape was open grassland. appears to lead by 50e70 years, a margin larger than the sub- Natural variability in d13C values was ~4.5‰ prior to the excursion decadal 2s errors for most of the nearest UeTh ages, suggesting and 4‰ after, both of which are typical of natural variability asso- that the difference is not due to age model errors. One explanation ciated with moderate fluctuations in soil respiration rates and could be recrystallization of stalagmite AB2, which would lead to airewater interactions in the epikarst. The steep, replicable, d13C uranium loss and anomalously older ages. Evidence for this increase, with no associated d18O increase, indicates that the includes anomalously low uranium concentrations in two of the change in dominant vegetation occurred independently of climatic three nearest ages in AB2 (1e2 ppm vs. 5e13 ppm in the remainder change. of the stalagmite). In addition, Scroxton et al. (2018) documented The d18O records from Anjohibe stalagmites are interpreted as that portions of AB2 have been recrystallized from aragonite to proxies for past strength of the northwest Madagascar monsoon, calcite. Higher uranium concentrations in the stalagmite elsewhere likely correlated to precipitation amount in the region. No long- suggest that the AB2 age model remains robust at other depths. term trend or monotonic shift is observed in the d18O records, Therefore, we conclude that the land-use transition above indicating that no substantial drying occurred over the last 1700 Anjohibe started no earlier than 830 CE, but more likely began years (the length of the records). On a centennial scale, however, around 900 CE. Similarly, the modest drying began no earlier than some centuries were modestly wetter or drier than others. Using 930 CE but more likely around 1000 CE. Crucially, as both the Bayesian Change Point Analysis of the high resolution d18O record vegetation and climate proxies come from different depths in the from stalagmite AB2 (Fig. 4f), Scroxton et al. (2017) documented same stalagmites, the evidence for (modest) drying occurs after the fluctuations in wet and dry conditions every ~500 years over the d13C change and associated land-use change. last 1700 years. Notably, there was a decrease in stalagmite d18O The climate and vegetation signals generated by stalagmites values indicating a period of relatively wet conditions between 480 from a single cave such as Anjohibe cannot necessarily be extrap- and 960 CE, with the wettest conditions of the last 1700 years be- olated to represent conditions everywhere on Madagascar. tween 780 and 960 CE. This wetter period was followed by However, there are several reasons to infer that the results from the increased stalagmite d18O values between 960 and 1485 CE, which stalagmites are indicative for much of the island. For one, sedi- we interpret as reduced precipitation. This transition from wet to ments from lakes across Madagascar, including Kavitaha, Tri- dry is identified at 960 CE by Bayesian Change Point analysis but is trivakely, Mitsinjo and Amparihibe, show increases in charcoal and/ unlikely to have been instantaneous. A period of climatic transition or Gramineae pollen percentages sometime between 700 and 1100 between 930 and 1000 CE is supported by the speleothem d18O CE (Burney, 1987a,b; Matsumoto and Burney, 1994; Gasse and Van record. Campo, 1998; Burney et al., 2004). The lake records do not have the 132 L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140

Figure 3. Map showing locations of sites with butchered subfossil vertebrate bones that fall within our search interval (50 BCE to 1350 CE). Butchered taxa and dates are indicated. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article). precise chronological control of the speleothem records, but Anjohibe is located near the northwest coast of Madagascar and its provide evidence of a similar landscape change over a broad area. yearly rainfall is dominated by the summer monsoon season, d18O Also, under monsoon systems, stalagmite d18O values are typi- values in AB2 should reflect the history of rainfall variability in the cally better indicators of regional rainfall than local precipitation southwestern Indian Ocean during the austral summer (or north- because the precipitation d18O values typically record convective western Madagascan) monsoon. Analysis of instrumental records processes rather than the point location rainfall amount (Risi et al., using data from Harris et al. (2014) supports the hypothesis that the 2008; Kurita et al., 2009; LeGrande and Schmidt, 2009). As wet and dry climate phases observed at Anjohibe were regional in L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 133

Table 2 Dated human-modified subfossil bones within our search interval (50 BCE to 1350 CE), ordered by date (14C dates are listed here are uncalibrated; CE dates are based on calibrated 14C dates).

14C yr BP ± SD Yr CE ± SD Lab # Element Spec # Taxon Sitea Sourceb

915 ± 35 1140 ± 120 CAMS 147113 Femur TAO-66-25 Propithecus verreauxi 11 940 ± 40 1130 ± 90 CAMS 147114 Femur TAO-66-24 Propithecus verreauxi 11 1045 ± 30 1065 ± 75 CAMS 143128 Tibia TAO-66-50 Propithecus verreauxi 12 1035 ± 25 1005 ± 135 CAMS 147111 Humerus TAO-66-2 Propithecus verreauxi 11 1055 ± 30 1000 ± 140 CAMS 147628 Femur TAO-66-23 Propithecus verreauxi 11 1130 ± 30 955 ± 65 CAMS 147036 Femur TAO-66-21 Propithecus verreauxi 11 1130 ± 30 955 ± 65 CAMS 147035 Femur TAO-66-FemA Propithecus verreauxi 11 1155 ± 25 940 ± 50 CAMS 147117 Femur TAO-66-FemC Propithecus verreauxi 11 1160 ± 25 940 ± 50 CAMS 147116 Femur TAO-66-5 Propithecus verreauxi 11 1150 ± 15 940 ± 50 UCIAMS 159131 Humerus UA 3629 HPL 13 Pachylemur insignis 22 1165 ± 25 935 ± 55 CAMS 147115 Femur TAO-66-FemB Propithecus verreauxi 11 1165 ± 30 930 ± 40 CAMS 147332 Femur TAO-66-30 Propithecus verreauxi 11 1215 ± 20 870 ± 95 UCIAMS 158554 Femur UA 3093 FPL 163 Pachylemur insignis 22 1230 ± 25 835 ± 65 UCIAMS 158553 Femur UA 3088 FPL 192 Pachylemur insignis 22 1245 ± 20 830 ± 55 UCIAMS 158552 Femur UA 3047 FPL 80 Pachylemur insignis 22 1240 ± 15 830 ± 55 UCIAMS 159129 Humerus UA 3610 HPL 100 Pachylemur insignis 22 1255 ± 15 825 ± 55 UCIAMS 159132 Humerus UA 3695 HPL87 Pachylemur insignis 22 1265 ± 20 825 ± 60 UCIAMS 158555 Femur UA 3133 FPL 74 Pachylemur insignis 22 1250 ± 15 825 ± 55 UCIAMS 159130 Humerus UA 3619 HPL 24 Pachylemur insignis 22 1295 ± 25 780 ± 90 CAMS 142604 Femur UA 1451 FAL 117 Archaeolemur majori 33 1297 ± 24 774.5 ± 98 OxA-33535 Tarsometatarsus MNHN MAD 6662 Mullerornis sp. ? 4 1296 ± 32 772.5 ± 93 UBA-19725 Tibiotarsus MNHN MAD 1906-16 Aepyornis maximus 44 1740 ± 50 385 ± 145 TO-1437 Femur MNHN MAD 1709 Hippopotamus lemerlei 55 1970 ± 50 100 ± 130 AA 2895 Femur MNHN MAD 1711 Hippopotamus lemerlei 45

a Site: 1 Taolambiby, 2 Tsirave, 3 Manombo, 4 Ambolisatra, 5 Lamboharana. ¼ ¼ ¼ ¼ ¼ b Source: 1 Crowley and Godfrey (2013);2 Crowley et al. (2017);3 Crowley (2010);4 Hansford et al. (2018);5 MacPhee and Burney (1991). ¼ ¼ ¼ ¼ ¼ scope. Rainfall variability at Anjohibe from 1901 to 2014 is signifi- conditions would have been similar to today. Testing this theory, cantly correlated with rainfall across much of Madagascar including and whether or not it applies during times of substantially different the northwest, north-central, north, part of the west, and even part climate regimes such as the last glacial period, is an area of ongoing of the east (14 to 18S, 43 to 48E) (Fig. 5). The region of high research. correlation encompasses four World Meteorological Organization Further afield, records of centennial rainfall variation in East stations and so is not a point source of correlation but a real field. Africa over the last 2000 years also match those for northwestern Outside of this field, the d18O values of stalagmite AB2 may record a Madagascar (Scroxton et al., 2017). At Lake Challa (on the borders proportion of rainfall variability. As AB2 d18O likely records past of Kenya and Tanzania), a wet period beginning 650 CE and ending variability in the northwestern monsoon, the regional correlations at 950 CE was followed by an extended dry period between 1170 outlined above allow us to assume this record likely reflects rainfall and 1300 CE, with drying conditions between the two (Buckles in other areas where the great majority of annual rainfall derives et al., 2016). The major wet and dry phases of the climate re- from the northwestern monsoon. Consequently, the d18O record cords of northwestern Madagascar also broadly correlate with from Anjohibe may help explain a substantial amount of rainfall climate records at Lake Naivasha, Kenya, which experienced a dry variability in other parts of Madagascar where monsoonal rainfall is period beginning around 1000 CE and lasting until 1270 CE significant, although it cannot be expected to capture variability in (Verschuren et al., 2000). This dry period followed a wet period of alternate moisture sources. In effect, as the proportion of unspecified duration; records for Lake Naivasha begin only around monsoonal rainfall to total rainfall decreases, the ability for the 1100 years ago. The broad similarities between these three sites stalagmite d18O values from Anjohibe to account for past rainfall may relate to the fact that East Africa and northwestern variability also decreases proportionately. Madagascar share a dominant source of moisture, i.e., the equa- Of particular interest is southwest Madagascar, as virtually all of torial West Indian Ocean. our butchery data and a substantial portion of the subfossil chro- nology records derive from this part of the island. Due to the rain- 4. Discussion shadow effect of the Madagascan highlands blocking easterly trade wind moisture, rainfall along the entire west coast is dominated by 4.1. Chronological overview austral summer moisture from the northwestern monsoon. Therefore, while southwest Madagascar is substantially drier than We can now summarize chronological data for megafaunal the northwest, the vast majority of its rain nevertheless falls during decline, butchery of wild animals, climate change, and vegetational the northwestern monsoon. Local conditions such as sea-surface shifts in Madagascar. Specifically we can compare the trajectory of temperatures may locally modulate rainfall (Zinke et al., 2005). megafaunal decline (as summarized using maximum likelihood c2 The Anjohibe speleothem d18O record agrees well with the statistics) and the trajectory of human population growth on Itafy Reef/Tulear Coral SST reconstruction from 1660 to 1975 CE Madagascar (derived from human genetic studies, Pierron et al., (compare Scroxton et al., 2017 and Zinke et al., 2014). This suggests 2017) to: (1) the shift in wild lemur prey targeted by human that local and regional oceanic and atmospheric conditions have hunters from mostly extinct to mostly extant species; (2) the record acted upon rainfall in the same direction during at least the last 350 of vegetation change from C3 to C4; and (3) the record of change in years, and therefore that the speleothem d18O record of monsoonal rainfall amount in the northwest, measured at close to annual variability from Anjohibe is applicable to the fossil sites of the resolution and showing centennial scale variability. With these southwest, at least during the late Holocene when baseline climatic comparisons, we can address our main questions regarding climate 134 L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140

Figure 5. Spatial correlation of rainfall anomalies between Anjohibe and a regional 0.5 grid. Data were collected from the CRU TS3.23 dataset (Harris et al., 2014) using the hydrological year (JulyeJune) from 1901 to 2014, and assessed using the KMNI climate explorer (Trouet and Van Oldenborgh, 2013). Colored regions have correlations significant at p < 0.05. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article).

centuries CE in the northwest, they also show that extreme drought did not occur during that time. There was no major, island-wide drought at around 950 years ago. Furthermore, the megafaunal

Figure 4. Speleothem evidence of vegetation changes and paleoclimate transitions crash pre-dated this period, as did the transition in the northwest between 500 and 1200 CE. a) Stalagmite AB3 ages with ±2 standard deviation error; from a C3- to a C4-dominated landscape. Indeed, that transition b) Stalagmite AB3 d13C values; c) Stalagmite AB3 d18O values; d) Stalagmite AB2 ages occurred almost entirely during an unusually wet period. Finally, no 13 with ±2 standard deviation error; e) Stalagmite AB2 d C values; f) Stalagmite AB2 trend toward increasing aridification is recorded in the stalagmite 18 d O values, with Bayesian Change Point Analysis modeled mean (dark blue line); g) climate records of northwestern Madagascar over the past 1700 Stalagmite MA3 ages with ±2 standard deviation error; h) Stalagmite MA3 d13C 18 years. values; i) Stalagmite MA3 d O values. Yellow box highlights transition from C3 to C4 vegetation. The synergy hypothesis fares a bit better. According to this hy- pothesis, at any given location, megafaunal decline preceded epi- sodes of habitat transformation; in effect, one can infer that the decline of the megafauna sparked habitat loss rather than the other variability and human activities as they do or do not relate to way around. Our data lend some support for this inference, as ev- megafaunal decline. idence of a hunting economy pre-dates the expansion of grasslands It is clear from the data described above and pictorially repre- in the northwest and elsewhere. However, the synergy hypothesis sented in Figure 6 that the greatest change in extinct-to-extant holds that the megafaunal crash occurred within a few centuries of vertebrate odds at subfossil sites (Fig. 6a) occurred between 700 human arrival and well prior to the introduction of cattle and and 900 CE, just as the human population of Madagascar began to expansion of agropastoralism. This argument in turn depends on expand markedly (Fig. 6b) (Pierron et al., 2017). Towards the end of evidence of humans arriving only a few centuries prior to the start that 200-year period of rapid megafaunal decline, human hunters of the 1st millennium, megafauna declining early in the 1st mil- in southwest Madagascar shifted from targeting now-extinct to lennium, and cattle arriving late in the 1st millennium. Support for still-extant species (Fig. 6c). Simultaneously, there was a significant this chronology was derived from research on megafaunal butchery shift in the northwest from a landscape dominated by C3 to C4 at Taolambiby in south central Madagascar, and the abundance of plants, likely signaling a local economic change from hunting/ spores of a mega-dung fungus called Sporormiella ( Preussia spp., ¼ gathering to more dedicated agropastoralism (Fig. 6d). All of this see Kruys and Wedin, 2009) in sediments at four sites: Ambolisatra preceded a moderate change in rainfall amount from wet to dry in and Belo-sur-Mer on the west coast, Amparihibe in the northwest, the same region (Fig. 6e). and Kavitaha in central Madagascar (Burney et al., 2003, 2004; This chronology challenges both the aridification and the syn- Perez et al., 2003, 2005). Because of the life history requirements ergy hypotheses. With regard to the aridification hypothesis, while of Sporormiella, counts of its spores in stratigraphic columns at our data support dry conditions during the late 10th and early 11th various sites have been used as proxies for large-animal abundance. L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 135

and Kavitaha almost 1000 years later as evidence of the introduc- tion of cattle. However, radiocarbon records on Madagascar's sub- fossils across the island support rapid megafaunal loss coeval with spikes in Sporormiella spores toward the end of the 1st millennium. There is excellent evidence that humans arrived much earlier, indeed, in the early Holocene. Hansford et al. (2018) report mega- faunal butchery during the early and middle Holocene at several sites in the interior of Madagascar. These dates were derived from elephant bird specimens showing clear and purposeful perimortem damage; the oldest such records (at Christmas River, south central Madagascar) exceed 10,000 years. These dates match dates for unmodified bones from the same stratigraphic layer (Muldoon et al., 2012). Butchered Mullerornis at Lamboharana on the west coast dates to over 6000 years (Hansford et al., 2018). Interestingly, two bird species, long believed to have been introduced to Madagascar by early human settlers (Hawkins and Goodman, 2003; Blench, 2008), are now confirmed to have been present on Madagascar in the late Pleistocene or early Holocene (Goodman et al., 2013). Early radiocarbon dates at Ankilitelo for the helmeted guineafowl (Numida meleagris at >13,000 years agp) and the pied crow (Corvus albus, at 9000 years agp) were taken by Goodman et al. (2013) as evidence of natural dispersal from Africa, a possibility that cannot be ruled out. However, guineafowl are virtually flightless, and easily transported as food on boats; wild corvids are known to travel as stow-aways on ships, and the Mal- agasy pied crow is morphologically identical to members of the same species in Africa. Hansford et al.'s (2018) dates for early megafaunal butchery lend strong support for a very early human colonization of the island of Madagascar, and make plausible the notion that >13,000 year-old helmeted guineafowl (N. meleagris) and >9000 year-old pied crow (C. albus) at Ankilitelo were intro- duced by humans. Pierron et al. (2017) found no genetic evidence of such an early human colonization. However, they could not exclude the existence of small early populations effectively leaving no genetic signature. They did identify two splits of proto-Malagasy populations from Austronesian and southern African Bantu relatives, the first from south Borneo between 3000 and 2000 years ago, and the second from Africa around 1500 years ago. Recognizing the need to ac- count for the potential incompleteness of their comparative Austronesian and African samples, they also stated that divergence could have been earlier. In other words, there is a possibility that DNA from descendent populations from the common ancestor population at the time of the earliest split was not included in the analysis, or that such a population no longer exists in the place of origin. This implies a minimum age of population divergence be- tween 2000 and 3000 years ago. As stated by Pierron et al. (2017: p. E6502), “…these dates reflect the age of the oldest possible common ancestors between Malagasy and the African/Indonesian Figure 6. Evidence of changes in extinct:extant animal odds, vegetation, climate, and sampled populations, meaning that the departure to Madagascar is human population between 500 and 1200 CE. a) Plot of maximum likelihood c2 sta- not later but could be earlier than these dates”. tistics for comparisons with different cut-points, showing inversion in extinct:extant There are additional records of human presence on Madagascar subfossil odds at ~850 CE; b) Estimation of changes in the effective population size more than 2000 years ago. For example, Douglass (2016) reports across time, using genome-wide identity-by-descent (IBD) sharing (Pierron et al., 2017); c) Radiocarbon ages of subfossil bones with butchery marks, showing extinct the existence of marine shell middens from the coastal site of species (filled circles) and extant species (open circles), with error bars of ±1 standard Velondriake in southwest Madagascar; a worked marine shell from deviation; d) Speleothem d13C values from stalagmites AB3, AB2 and MA3 from this locality has yielded a calibrated age between 825 and 504 BCE. Anjohibe Cave (green lines) (Burns et al., 2016; Scroxton et al., 2017; Voarintsoa et al., In addition, optically simulated luminescence (OSL) dates on quartz 2017); e) Speleothem d18O from stalagmite AB2 from Anjohibe Cave (blue), with mean 18 grains associated with stone artifacts (microliths) at Lakaton'i Anja d O values as modeled using Bayesian Change Point Analysis (dark blue line). In all panels, shading indicates most likely period of change. Cave in northwest Madagascar place foragers on this part of the island more than 4000 years ago (Dewar et al., 2013). These dates were rejected by Anderson et al. (2018) on the basis of discrep- Burney et al. (2003) interpreted drops in Sporormiella spores at ancies between the OSL dates and much younger 14C dates for Ambolisatra and Belo-sur-Mer during the early part of the 1st charcoal in the same stratigraphic layers. However, Dewar et al. millennium as signals of decimation of megafaunal populations, (2013) explained the contradictory dates as resulting from ter- and they interpreted spikes in Sporormiella spores at Amparihibe mites causing downward migration of younger charcoal via 136 L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 bioturbation. Dewar et al. (2013) noted that the OSL dates are in their geographic origin is uncertain. Arabian traders established stratigraphic order and the 14C dates are not; the latter are clearly large trading posts in Madagascar and India (Chaudhuri, 1985), disturbed. Termites can be expected to deliberately move fresh and while Indonesians established a trade network that connected charred wood into lower strata for food; charcoal is also trans- southeast Asia with Madagascar, East Africa and the Swahili ported downward by termites for underground moisture control. city-states (Boivin et al., 2013). Termites would not deliberately move rocks. Thus, Dewar et al. (2013) inferred that the stone artifacts in the lower layers are 4.2. The subsistence shift hypothesis indeed in situ and over 4000 years old. The counterargument that these artifacts are also intrusive must depend on accidental mixing. We offer the “Subsistence Shift Hypothesis” to explain mega- There is no evidence, furthermore, that the tools in the lower layers faunal extinction in Madagascar. This hypothesis maintains that the are coeval with introduced crops (Pomerantz, 2017), which do not trigger for rapid megafaunal decline in Madagascar was the expan- appear at this site until after 1000 CE (Crowther et al., 2016). Finally, sion of the Indian Ocean trade network and the concomitant tran- bones with cut marks have been found at Lakaton'i Anja Cave, but sition from an economy that depended largely on hunting and have not been described (Ekblom et al., 2016). foraging to one that relied increasingly on agropastoralism, trade It is curious that Anderson et al. (2018) rejected early human and urbanism. Megafaunal decline intensified as Madagascar butchery for the nearly-2000-year-old hippopotamus bones from received an influx of settlers bringing new crop plants and domes- Ambolisatra and Lamboharana that MacPhee and Burney (1991) ticated animals. reported as anthropogenically modified. While accepting anthro- The subsistence shift hypothesis is supported by the fact that the pogenic modification for these bones, Anderson et al. (2018) argued megafaunal crash was not coincident with initial human arrival, that they may have been damaged by excavators or museum pre- nor did it follow human arrival within a few centuries. It was not parators. These bones were collected in southwest Madagascar by triggered by aridification. It did, however, begin at around the time Guillaume Grandidier at the turn of the twentieth century (e.g., cattle were introduced to Madagascar, and it peaked as human Grandidier, 1900; see summary by Chanudet, 1975). Ambolisatra populations began to grow rapidly. The availability of alternative was a marsh at that time; sometimes collectors had to wade into foods (such as cattle, goats, sheep) did not relieve pressure on the water that was neck deep and dive under water to retrieve sub- megafauna, which increased as human populations grew until the fossils. When Grandidier sent these bones to Paris, he reported that large-bodied endemic animal species could no longer maintain some showed perimortem human modification. To confirm this, viable populations. Megafaunal hunting then declined in economic MacPhee and Burney (1991) ran chopping experiments on subfossil importance not because alternative introduced foods were now hippo bone from Ampoza as well as fresh cow bone, and found that available, but because the megafaunal populations themselves had the subfossil bone splintered, while fresh bone did not. It was crashed. As Dewar et al. (2013: 12587) observed, “The activities of precisely for this reason that MacPhee and Burney (1991) rejected foraging [human] populations have environmental consequences excavation damage as a viable explanation for the chops they that differ in both degree and nature from those of Iron Age farmers documented on Grandidier's hippo bones from Lamboharana and and pastoralists, and changes in paleoenvironmental proxies Ambolisatra. interpreted as signaling ‘human arrival’ may in fact be signals of a Pierron et al. (2017) provide evidence of human population change in human economy.” admixture and a rapid expansion of the effective breeding popu- Our hypothesis does not deny that there were dramatic fluctu- lation of humans well after initial human arrival on Madagascar. ations in climate during the late Pleistocene and Holocene, or that That expansion was coincident with archaeological evidence for a these fluctuations resulted in local changes in species distributions. changing economy, from one that depended primarily on hunting/ Nor does it deny that, prior to the subsistence shift described here, foraging to one that depended much more on herding/farming hunters and foragers may have had negative impacts on Mada- (Table 3). It began in the north around 700 CE ( 1250 BP), and then gascar's megafauna. It does not preclude the notion that natural ¼ spread to other parts of Madagascar (Pierron et al., 2017). Archae- fires would have increased in frequency and intensity as mega- ological evidence of intensified forest resource extraction and faunal populations declined. We do argue, however, that human clearing of land appeared along the east coast in the 7th century CE arrival, by itself, does not trigger extinction; rather, the degree to and extended through the 11th century (Agarwal et al., 2005). which terrestrial vertebrate species are threatened depends on The expansion of the human breeding population was linked to human population size and human subsistence strategy, as well as a growing Indian Ocean trading network that connected the reproductive dynamics of the targeted species. The expansion Madagascar, China, India, Indonesia, Arabia, and Africa (Blench, of Madagascar's trade network, influx of new settlers, and eco- 2008; Fuller and Boivin, 2009; Boivin et al., 2013; Dewar, 2014; nomic transition from hunting and foraging to herding and farming Douglass and Zinke, 2015; Radimilahy and Crossland, 2015; triggered rapid growth in the size of the effective human breeding Crowther et al., 2016; Ekblom et al., 2016; Pomerantz, 2017; population, greater direct exploitation of wild (in addition to Prendergast et al., 2017). From ~700 CE onward, Madagascar domesticated) animals, and more rapid habitat modification (a experienced an influx of wild and domesticated plants and animals change in the fire ecology of the island). Prior to this, small human from continental Africa and from Asia (Dewar, 2014). Iron tools populations appear to have continually or intermittently occupied became more central to the economy (Dewar, 2014). Villages and Madagascar over an extended period of time without triggering hamlets spread throughout the island. Between 700 and 1200 CE, extinction. Coastal populations likely exploited marine food re- waves of Sama-Bajaw Austronesians brought crops, such as co- sources while people living in the interior relied more on endemic conuts (Cocos), yams (Dioscoria), rice (Oryza), saffron (Crocus) and terrestrial animals and wild plants. Megafaunal extinction occurred cotton (Gossypium), to moist and highland habitats (Crowther et al., only after the expansion of the Indian Ocean trade network. 2016; Beaujard, 2017). Bantu people, mostly from the Swahili/ Following that expansion, coastal people may have continued to Sabaki group, brought cattle (Bos), goats (Capra), chickens (Gallus), exploit marine resources, minimally impacting terrestrial verte- and perhaps wild boars (Potamochoerus larvatus)(Blench, 2008), as brates, but forest clearance through cutting and burning would well as plants adapted to thrive in dry habitats, including sorghum have served the economic needs of both herders and farmers, who (Sorghum), groundnuts (Vigna), and taro (Colocasia)(Beaujard, would have also hunted wild animals to varying degrees (as is true 2017). Bananas (Musa acuminata) also arrived early, although today). L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 137

Table 3 Subsistence transition illustrated by plants or animals recorded at archaeological sites, or possible human occupation zones at paleontological sites, within our search interval (50 BCE to 1350 CE).

Site Time (CE) Plants and Animals Sources

Ambolisatra, SW coastal 100 CE Hippopotamus butchery site MacPhee and y Burney (1991) Lamboharana, SW coastal 385 CE Hippopotamus butchery site MacPhee and y Burney (1991) Lakaton'i Anja (Andavakoera 5th to 10th centuries CE Mollusks Dewar and Gorge), N Coastal Wild terrestrial animals Rakotovololona (1992) Velondriake, SW coastal, Phase I occupation begins ~550 CE and ends ~1050 CE Fish (including Emperors, Breams, Unicorn fish, Douglass et al. Antsaragnagnangy (ANGY I) Parrotfish, Groupers, and Rabbitfish) (2018) Terrapins Turtles Tortoises Frogs Birds (including *Gallus gallus or *Numida meleagris) Bats Tenrecs Small and medium-sized lemurs Sarodrano, SW coastal Beginning in the mid-1st millennium *Bos Battistini and Verin Fish (1971) Andrianaivoarivony (1987) Irodo, NE coastal 7th century *Bos Andrianaivoarivony Fish (1987) West Mikoboka Plateau, cave 710 CE *Rattus rattus Crowley et al. (2017) #13, SW Ambolisatra, SW coastal 770 CE Aepyornis maximus butchery site Hansford et al. y (2018) Manombo Toliara, SW coastal 780 CE Archaeolemur majori butchery site This paper y Ampasimahavelona NE coastal 8th to 10th centuries CE Shell midden Crowther et al. *Musa acuminata (2016),a *Oryza sativa Pomerantz (2017) Ampasimaiky Rock Shelter, SW Dating uncertain; possibly associated stone tools *Rock art depictions of humped cattle Rasolondrainy interior (2012) Tsirave, SC interior 825-940 CE Archaeolemur and Pachylemur insignis This paper y y butchery site Amparihibe, NW coastal 780e1010 CE (single date, 2 SD) *Proliferation of cattle inferred from increase in Burney et al. (2003) Sporormiella spore count Anjohibe, NW Beginning 890 CE, over a period of 100 years *Proliferation of cattle inferred from intense Burns et al. (2016) burning episode recorded in stalagmites Mahilaka, NW coastal Beginning in the late 1st millennium *Bos Radimilahy (1998) *Capra Crowther et al. *Ovis? (2016) *Gallus gallus Pomerantz (2017) *Potamochoerus larvatus Prendergast et al. *Rattus rattus (2017) *Mus musculus *Musa acuminata *Oryza sativa *Gossypium arboretum Tsiandrora, Maliovola, Mokala, Possibly beginning in the late 9th century; mostly early 2nd *Bos (at Tsiandrora) Rakotoarisoa (1998) Ambinanibe, Ndrenany, SE millennium, from the 11th through the 13th centuries *Oryza sativa (at Maliovola) *Colocasia esculenta (at Maliovola) Fish Shellfish Small wild animals (hunted) Kavitaha, CH interior 900e1260 CE (single date, 2 SD) *Proliferation of cattle inferred from increase in Burney et al. (2003) Sporormiella spore count Taolambiby, SW interior 930e1140 CE Propithecus verreauxi butchery site This paper Andaro, S 10th e 13th centuries CE *Bos Parker-Pearson et al. *Capra or Ovis (2010) *Gallus gallus *Potamochoerus larvatus Andranosoa, S Beginning in the 10th century CE *Bos Rasamuel (1984) *Ovis aries Andrianaivoarivony *Capra? (1987) *Felis cattus Rakotozafy and *Canis familiaris Goodman (2005) *Potamochoerus larvatus Boivin et al. (2013) Fish Shellfish Tenrec (continued on next page) 138 L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140

Table 3 (continued )

Site Time (CE) Plants and Animals Sources

Giant tortoise y Hippopotamusb y Lakaton'i Anja, Andavakoera 11th e 12th centuries CE Fish and shellfish Dewar and Gorge, upper layers, N coastal Giant tortoise Rakotovololona y Tenrecs (1992) Birds Crowther et al. Giant lemurs (2016) y *Bos *Capra or Ovis *Potamochoerus larvatus or possibly domesticated pig? *Gossypium arboretum Talaky, S coastal 11th e 13th centuries CE Mollusks Battistini et al. Aepyornis (eggshell) (1963) y Ankadivory D'Ralambo, CH 12th 14th centuries CE *Bos Rakotovololona À interior, NE of Antananarivo *Musa acuminata (1993) *Oryza sativa Pomerantz (2017) Lakaton'ny akanga, N coastal 13th century CE Subfossil (extinct animal) bones and bones of Dewar and y (Grotte des Pintades) other forest animals Rakotovololona Fish and shellfish (1992) Aepyornis (eggshell debris) y (No remains of domesticated animals) Rezoky, SC interior 13th e 15th centuries CE *Bos Verin and Battistini, *Capra hircus 1972 *Ovis aries Rakotozafy and *Canis familiaris Goodman (2005) *Potamochoerus larvatus Boivin et al. (2013) Tenrec ecaudatus Douglass et al. *Numida meleagris (2018) Mollusks Asambalahy, SC interior 14th century onward CE *Bovids Rakotozafy and Hippopotamus (undated, possibly cut-marked, Goodman (2005) y unverified) Boivin et al. (2013) *Numida meleagris *Canis familiaris

Extinct, *Introduced to Madagascar. y a Crowther et al. (2016) did not find charred food crop remains at Ampasimahavelona, but Pomerantz (2017) identified phytoliths of bananas and rice here. b Hippopotamus teeth from Andranosoa found in a garbage pit were illustrated and described as “giant subfossils,” but not identified as hippos by Rasamuel (1984).

The complete transformation of dry forested habitats to savannas as tenrecs, persist, despite the fact that they were hunted in the such as that seen at Anjohibe at 890 CE (Burns et al., 2016) cannot be past (their bones appear often in the middens of archaeological attributed to natural fire. This transformation was abrupt and it was sites), and continue to be hunted heavily in some areas today. complete in little more than a century, which strongly suggests Domesticated species such as cattle can sustain large human habitat modification by humans. Megafauna also disappeared from populations because individuals grow rapidly to large body size, wetter regions such as the subhumid Central Highlands and the sexually mature early (in less than two years), and have short rainforests of the eastern escarpment and lowlands. They dis- interbirth intervals (Marshall, 1989). appeared from habitats that were previously forested and from old woodland and savanna habitats. We suggest that megafauna were 5. Conclusions hunted and that fires were set by people in all habitats. However, in wetter areas, forest regrowth was rapid, and habitat transformation In summary, we are arguing that explanations for megafaunal was relatively slow. For this reason, megafauna disappeared from extinction must consider subsistence strategies and human popu- the eastern forests without wholesale destruction of the forested lation size. We suggest that, in Madagascar, a fundamental shift in habitats. The disappearance of the megafauna from all habitats how humans interacted with their natural environment occurred suggests that these animals were hunted everywhere, even as as trade connections expanded and reliance on agropastoralism habitat transformation varied across the island. and urbanism grew. Counterintuitively, and despite the availability Human populations tend to expand with increased agro- of alternative resources, hunting pressure on wild animals pastoralism because of increased carrying capacity; it is carrying increased with agropastoralism, impacting large animals in all capacity that controls the maximum human population size that habitats. Habitat modification through burning had a compound- can be supported by available resources (Bettinger, 2016; Zahid ing, negative effect on large-bodied endemic species. et al., 2016). In general, carrying capacity is much higher for Our more fundamental argument is that human presence is not plants than animals, for domesticated plants than wild plants, and automatically catastrophic, but that certain human activities can be for domesticated animals than wild animals. This is related to the catastrophic, and their effects can be entirely independent of reproductive resilience of the targeted food resource species, which climate change. We acknowledge that while our data on the in turn is affected by their density, growth rate and life history megafaunal crash are island-wide, our data on changes in rainfall characteristics (Winterhalder and Goland, 1993). Many of the large, derive mainly from the northwest, and our butchery data derive endemic animals of Madagascar were slow reproducers and from southwestern and south central Madagascar. It is critical that therefore vulnerable to extinction (Johnson, 2002; Richard et al., future research focus more precisely on the temporal and spatial 2002; Schwartz et al., 2002, 2007; Godfrey et al., 2006b; Catlett dynamics of human activities, climate variation and extinction in et al., 2010). Those endemic species that reproduce rapidly, such Madagascar (e.g., Douglass et al., 2018). Regional variation in L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140 139 subsistence norms and human population density must be a critical Chaudhuri, K.N., 1985. Trade and Civilization in the Indian Ocean: An Economic part of the discussion. History from the Rise of Islam to 1750. Cambridge University Press, Cambridge. Cheng, H., Lawrence Edwards, R., Edwards, R.L., Shen, C.-C., Polyak, V.J., Asmerom, Y., Woodhead, J., Hellstrom, J.C., Hellstrom, J., Wang, Y., Kong, X., € 230 Acknowledgements Spotl, C., Wang, X., Calvin Alexander Jr., E., 2013. Improvements in Th dating, 230Th and 234U half-life values, and UeTh isotopic measurements by multi- collector inductively coupled plasma mass spectrometry. Earth and Planetary The conclusions reported here were drawn from research con- Science Letters 371, 82e91. fi ducted under collaborative international research accords that have Crowley, B.E., 2010. A re ned chronology of prehistoric Madagascar and the demise of the megafauna. Quaternary Science Reviews 29, 2591e2603. supported the compilation of a large radiocarbon database and Crowley, B.E., Godfrey, L.R., 2013. Why all those spines? Anachronistic defenses in facilitated paleoclimatological research and cut mark analysis of the Didiereoideae against now extinct lemurs. South African Journal of Science fossils. We gratefully thank the government of Madagascar for 109, 70e76. Crowley, B.E., Samonds, K.E., 2013. Stable carbon isotope values confirm a recent granting us permission to collect fossils and stalagmites and to increase in grasslands in northwestern Madagascar. Holocene 23(7), analyze samples. This manuscript was greatly improved by 1066e1073. https://doi.org/10.1177/0959683613484675. comments from anonymous reviewers and editors, and by con- Crowley, B.E., Godfrey, L.R., Bankoff, R.J., Perry, G.H., Culleton, B.J., Kennett, D.J., Sutherland, M.R., Samonds, K.E., Burney, D.A., 2017. Island-wide aridity did not versations with Drs. Alison Richard, Kristina Douglass, and George trigger recent megafaunal extinctions in Madagascar. Ecography 40, 901e912. Perry. This work was supported by the National Science Foundation Crowther, A., Lucas, L., Helm, R., Horton, M., Shipton, C., Wright, H.T., Walshaw, S., [AGS-1702891 to Stephen J. Burns, BCS-1750598 to Laurie R. God- Pawlowicz, M., Radimilahy, C., Douka, K., Picornell-Gelabert, L., Fuller, D.Q., fi frey, AGS-1702691 to David McGee, and BCS-1749676 to Brooke E. Boivin, N.L., 2016. Ancient crops provide rst archaeological signature of the westward Austronesian expansion. Proceedings of the National Academy of Crowley]. Sciences of United States of America 113(24), 6635e6640. Dewar, R.E., 1984. Extinctions in Madagascar: The loss of the subfossil fauna. In: Martin, P.S., Klein, R.G. (Eds.), Pleistocene Extinctions: The Search for a Cause. References Yale University Press, New Haven, pp. 574e593. Dewar, R.E., 2014. Early human settlers and their impact on Madagascar's land- Agarwal, D.K., Silander Jr., J.A., Gelfand, A.E., Dewar, R.E., Mickelson Jr., J.G., 2005. scapes. In: Scales, I.R. (Ed.), Conservation and Environmental Management in Tropical deforestation in Madagascar: analysis using hierarchical, spatially Madagascar. Routledge, Taylor & Francis Group, London, pp. 44e64. explicit, Bayesian regression models. Ecological Modelling 185, 105e131. Dewar, R.E., Rakotovololona, S., 1992. La chasses aux subfossiles: les preuves de Anderson, A., Clark, G., Haberle, S., Higham, T., Nowak-Kemp, M., Prendergast, A., XIeme siecle au XIIIeme siecle. Taloha 11, 4e15. Radimilahy, C., Rakotozafy, L.M., Ramilisonina, Schwenninger, J.L., Virah- Dewar, R.E., Radimilahy, C., Wright, H.T., Jacobs, Z., Kelly, G.O., Berna, F., 2013. Stone Sawmy, M., Camens, A., 2018. New evidence of megafaunal bone damage in- tools and foraging in northern Madagascar challenge Holocene extinction dicates late colonization of Madagascar. PLoS One 13(10), e0204368. https:// models. Proceedings of the National Academy of Sciences of United States of doi.org/10.1371/journal.pone.0204368. America 110, 12583e12588. https://doi.org/10.1073/pnas.1306100110. Andrianaivoarivony, R., 1987. L'alimentation carnee chez les anciens malgaches. Dorale, J.A., Lawrence Edwards, R., Ito, E., Gonzalez, L.A., 1998. Climate and vege- Nouvelles du Centre d'Art et d'Archeologie 5e6, 22e26. tation history of the midcontinent from 75 to 25 ka: A speleothem record from Battistini, R., 1965. L'importance de l'action de l'homme dans les transformations Crevice Cave, Missouri, USA. Science 282, 1871e1874. proto-historiques du milieu naturel a Madagascar. Annales de l'Universite de Douglass, K., 2016. An archaeological investigation of settlement and resource Madagascar, Taloha 1, 215e223. exploitation patterns in the Velondiake Marine Protected Area, Southwest Battistini, R., Verin, P., 1971. Temoignages archeologiques sur la cote^ Vezo de Madagascar, ca. 800 BC to AD 1900. PhD Dissertation, Yale University. l'embouchure de l'Onilahy a la Baie des Assassins. Taloha 4, 51e63. Douglass, K., Zinke, J., 2015. Forging ahead by land and by sea: archaeology and Battistini, R., Verin, P., Rason, R., 1963. Le site archeologique de Talaky. Annales paleoclimate reconstruction in Madagascar. African Archaeological Review 32, Malgaches 1, 111e142. 267e299. Beaujard, P., 2017. Histoire et voyages des plantes cultivees a Madagascar avant le Douglass, K., Antonites, A.R., Quintana Morales, E.M., Grealy, A., Bunce, M., XVIe siecle. Karthala, Paris. Bruwer, C., Gough, C., 2018. Multi-analytical approach to zooarchaeological Bettinger, R.L., 2016. Commentary: Prehistoric hunter-gatherer population growth assemblages elucidates Late Holocene coastal lifeways in southwest rates rival those of agriculturalists. Proceedings of the National Academy of Madagascar. Quaternary International 471, 111e131. Sciences of United States of America 113(4), 812e814. Ekblom, A., Lane, P., Radimilahy, C., Rakotoarisoa, J.-A., Sinclair, P., Virah-Sawmy, M., Blench, R.M., 2008. The Austronesians in Madagascar and their interaction with the 2016. Migration and interaction between Madagascar and Eastern Africa, 500 Bantu of East African coast: surveying the linguistic evidence for domestic and BCEd1000 CE: An archaeological perspective. In: Campbell, G. (Ed.), Early Ex- translocated animals. Studies in Philippine Languages and Cultures 18, 18e43. change between Africa and the Wider Indian Ocean World. Palgrave Macmillan, Boivin, N., Crowther, A., Helm, R., Fuller, D.Q., 2013. East Africa and Madagascar in Cham, Switzerland, pp. 195e230. the Indian Ocean world. Journal of World Prehistory 26, 213e281. Fairchild, I.J., Baker, A., 2012. Speleothem Science: From Process to Past Environ- Brook, G.A., Rafter, M.A., Railsback, L.B., Sheen, S.-W., Lundberg, J., 1999. A high- ments, 1st ed. Blackwell Publishing Ltd., Oxford. resolution proxy record of rainfall and ENSO since AD 1550 from layering in Fuller, D.Q., Boivin, N., 2009. Crops, cattle and commensals across the Indian Ocean: stalagmites from Anjohibe Cave, Madagascar. Holocene 9, 695e705. Current and potential archaeobiological evidence. In: Lefevre, G. (Ed.), Plantes Buckles, L.K., Verschuren, D., Weijers, J.W.H., Cocquyt, C., Blaauw, M., Sinninghe et Societ es, Etudes Ocean Indien 42e43. CEROI-INALCO, Paris, pp. 3e46. Damste, J.S., 2016. Interannual and (multi-)decadal variability in the sedimen- Gasse, F., Van Campo, E., 1998. A 40,000 year pollen and diatom record from Lake tary BIT index of Lake Challa, East Africa, over the past 2200 years: assessment Tritrivakely, Madagascar, in southern tropics. Quaternary Research 49, 299e311. of the precipitation proxy. Climate of the Past 12, 1243e1262. Genty, D., Baker, A., Massault, M., Proctor, C., Gilmour, M., Pons-Branchu, E., Burgess, N., D'Amico Hales, J., Underwood, E., Dinerstein, E., 2004. Terrestrial Hamelin, B., 2001. Dead carbon in stalagmites: Carbonate bedrock paleo- Ecoregions of Africa and Madagascar. Island Press, Washington, D.C. dissolution vs. ageing of soil organic matter. Implications for C-13 variations in Burney, D.A., 1987a. Late Holocene vegetational change in Central Madagascar. speleothems. Geochimica et Cosmochimica Acta 65, 3443e3457. Quaternary Research 28(1), 130e143. Godfrey, L.R., Jungers, W.L., Burney, D.A., Vasey, N., Ramilisonina, Wheeler, W., Burney, D.A., 1987b. Late Quaternary stratigraphic charcoal records from Lemelin, P., Shapiro, L.J., Schwartz, G.T., King, S.J., Ramarolahy, M.F., Madagascar. Quaternary Research 28(2), 274e280. Raharivony, L.L., Randria, G.F.N., 2006a. New discoveries of skeletal elements of Burney, D.A., 1993. Late Holocene environmental changes in arid southwestern Hadropithecus stenognathus from Andrahomana Cave, southeastern Madagascar. Quaternary Research 40, 98e106. Madagascar. Journal of Human Evolution 51, 395e410. Burney, D.A., Robinson, G.S., Burney, L.P., 2003. Sporormiella and the late Holocene Godfrey, L.R., Schwartz, G.T., Samonds, K.E., Jungers, W.L., Catlett, K.K., 2006b. The extinction in Madagascar. Proceedings of the National Academy of Sciences of secrets of lemur teeth. Evolutionary Anthropology 15, 142e154. United States of America 100(19), 10800e10805. Gommery, D., Ramanivosoa, B., Faure, M., Guerin, C., Kerloc'h, P., Sen egas, F., Burney, D.A., Burney, L.P., Godfrey, L.R., Jungers, W.L., Goodman, S.M., Wright, H.T., Randrianantenaina, H., 2011. Les plus anciennes traces d'activites anthropiques Jull, A.J.T., 2004. A chronology for late prehistoric Madagascar. Journal of Human de Madagascar sur des ossements d'hippopotames^ subfossiles d'Anjohibe. Evolution 47, 25e63. Comptes Rendus Palevol 10, 271e278. Burns, S.J., Godfrey, L.R., Faina, P., McGee, D., Hardt, B., Ranivoharimanana, L., Goodman, S.M., Jungers, W.L., 2014. Extinct Madagascar: Picturing the Island's Past. Randrianasy, J., 2016. Rapid human-induced landscape transformation in University of Chicago Press, Chicago. Madagascar at the end of the first millennium of the Common Era. Quaternary Goodman, S.M., Rakotozafy, L.M.A., 1997. Subfossil birds from coastal locations in Science Reviews 134, 92e99. western and southwestern Madagascar: A paleoenvironmental reconstruction. Catlett, K.K., Schwartz, G.T., Godfrey, L.R., Jungers, W.L., 2010. “Life history space”:A In: Goodman, S.M., Patterson, B.D. (Eds.), Natural Change and Human Impact in multivariate analysis of life history variation in extant and extinct Malagasy Madagascar. Smithsonian Institution Press, Washington DC, pp. 257e279. lemurs. American Journal of Physical Anthropology 142(3), 391e404. Goodman, S.M., Raherilalao, M.J., Muldoon, K.M., 2013. Bird fossils from Ankilitelo Chanudet, C., 1975. Conditions geographiques et archeologiques de la disparition des Cave: Inference about Holocene environmental changes in Southwestern subfossiles a Madagascar. Master's Thesis, Universite de Bretagne Occidentale. Madagascar. Zootaxa 3750, 534e548. 140 L.R. Godfrey et al. / Journal of Human Evolution 130 (2019) 126e140

Grandidier, G., 1900. Note sur des ossements d'animaux disparus provenant Radimilahy, C.M., 1998. Mahilaka: An Archaeological Investigation of an Early Town d'Ambolisatra sur la cote^ SW de Madagascar. Bulletin du Museum National in Northwestern Madagascar. Uppsala University, Uppsala. d'Histoire Naturelle 214e218. Radimilahy, C.M., Crossland, Z., 2015. Situating Madagascar: Indian Ocean dynamics Hansford, J., Wright, P.C., Rasoamiaramanana, A., Perez, V.R., Godfrey, L.R., and archaeological histories. Azania 50, 495e518. Errickson, D., Thompson, T., Turvey, S.T., 2018. Early Holocene human presence Raison, J.P., Verin, P., 1967. Le site de subfossiles de Taolambiby (Sud-Ouest de in Madagascar evidenced by exploitation of avian megafauna. Science Advances Madagascar). Doit-il etre^ attribue a une intervention humaine? Annales de 4(9) eaat6925. l'Universite de Madagascar 7, 133e142. Harris, I., Jones, P.D., Osborn, T.J., Lister, D.H., 2014. Updated high-resolution grids of Rakotoarisoa, J.A., 1998. Mille Ans d'Occupation Humaine dans le Sud-Est de monthly climatic observations e the CRU TS3.10 Dataset. International Journal Madagascar: Anosy, Une Ile au Milieu des Terres. L'Harmattan, Paris. of Climatology 34, 623e642. Rakotovololona, H.F.S., 1993. Ankadivory et la Periode Fiekena: Debut d'Urbanisation Hawkins, A.F.A., Goodman, S.M., 2003. Introduction to the birds. In: Goodman, S.M., a Madagascar. Donnees Archeologiques sur l'Origine des Villes a Madagascar- Benstead, B.D. (Eds.), Natural Change and Human Impact in Madagascar. Mombassa. Musee d'Art et d'Archeologie, Antananarivo, Madagascar. Smithsonian Institution Press, Washington DC, pp. 1019e1044. Rakotozafy, L.M.A., Goodman, S.M., 2005. Contribution a l'etude zooarcheologique Hellstrom, J., 2003. Rapid and accurate U/Th dating using parallel ion-counting de la region du Sud-ouest et extreme^ Sud de Madagascar sur la base des multi-collector ICP-MS. Journal of Analytical Atomic Spectrometry 18, collections de l'ICMAA de l'Universite d'Antananarivo. Taloha 14e15, 10e20. 1346e1351. Rasamuel, D., 1984. Alimentation et techniques anciennes dans le sud malgache a Hixon, S.W., Elliot Smith, E.A., Crowley, B.E., Perry, G.H., Randrianasy, J., travers une fosse a ordures du XIe siecle. Etudes Ocean Indien 4(4), 81e109. Ranaivoarisoa, J.F., Kennett, D.J., Newsome, S.D., 2018. Nitrogen isotope (d15N) Rasolondrainy, T., 2012. Discovery of rock paintings and Libyco-Berber inscription patterns for amino acids in lemur bones are inconsistent with aridity driving from the Upper Onilahy, Isalo Region, Southwestern Madagascar. In: Chami, F., megafaunal extinction in south-western Madagascar. Journal of Quaternary Radimilahy, C. (Eds.), Studies in the African Past, vol. 10, pp. 173e195. Science 33, 958e968. Richard, A.F., Dewar, R.E., Schwartz, M., Ratsirarson, J., 2002. Life in the slow lane? Johnson, C.N., 2002. Determinants of loss of mammal species during the Late Demography and life histories of male and female sifaka (Propithecus verreauxi Quaternary ‘megafauna’ extinctions: life history and ecology, but not body size. verreauxi). Journal of Zoology 256, 421e436. Proceedings of the Royal Society B: Biological Sciences 269, 2221e2227. Risi, C., Bony, S., Vimeux, F., 2008. Influence of convective processes on the isotopic Kruys, A., Wedin, M., 2009. Phylogenetic relationships and an assessment of composition (d18O and dD) of precipitation and water vapor in the tropics: 2. traditionally used taxonomic characters in the Sporormiaceae (Pleosporales, Physical interpretation of the amount effect. Journal of Geophysical Research Dothideomycetes, Ascomycota), utilising multigene phylogenies. Systematics and 113 article D19306. Biodiversity 7, 465e478. Ruggieri, E., 2013. A Bayesian approach to detecting change points in climatic re- Kurita, N., Ichiyanagi, K., Matsumoto, J., Yamanaka, M.D., Ohata, T., 2009. The rela- cords. International Journal of Climatology 33, 520e528. https://doi.org/ tionship between the isotopic content of precipitation and the precipitation 10.1002/joc.3447. amount in tropical regions. Journal of Geochemical Exploration 102, 113e122. Salmona, J., Heller, R., Quem er e, E., Chikki, L., 2017. Climate change and human https://doi.org/10.1016/j.gexplo.2009.03.002. colonization triggered habitat loss and fragmentation in Madagascar. Molecular Lachniet, M.S., 2009. Climatic and environmental controls on speleothem oxygen- Ecology 26, 5203e5222. isotope values. Quaternary Science Reviews 28, 412e432. https://doi.org/ Schwartz, G.T., Samonds, K.E., Godfrey, L.R., Jungers, W.L., Simons, E.L., 2002. Dental 10.1016/j.quascirev.2008.10.021. microstructure and life history in subfossil Malagasy lemurs. Proceedings of the LeGrande, A.N., Schmidt, G.A., 2009. Sources of Holocene variability of oxygen National Academy of Sciences of United States of America 99(9), 6124e6129. isotopes in paleoclimate archives. Climate of the Past 5, 441e455. https:// Schwartz, G.T., Godfrey, L.R., Mahoney, P., 2007. Inferring primate growth, devel- doi.org/10.5194/cp-5-441-2009. opment and life history from dental microstructure: The case of the extinct MacPhee, R.D.E., Burney, D.A., 1991. Dating of modified femora of extinct dwarf Malagasy lemur, Megaladapis. In: Bailey, S.E., Hublin, J.J. (Eds.), Dental Per- hippopotamus from southern Madagascar: implications for constraining hu- spectives on Human Evolution: State of the Art Research in Dental Paleoan- man colonization and vertebrate extinction events. Journal of Archaeological thropology. Springer, New York, pp. 147e162. Science 18, 695e706. Scroxton, N., Burns, S.J., McGee, D., Hardt, B., Godfrey, L.R., Ranivoharimanana, L., Mahe, J., Sourdat, M., 1972. Sur l'extinction des vertebr es subfossiles et l'aridifica- Faina, P., 2017. Hemispherically in-phase precipitation variability over the last tion du climat dans le Sud-Ouest de Madagascar. Bulletin de la Societ e 1700 years in a Madagascar speleothem record. Quaternary Science Reviews Geologique de France 1, 295e309. 164, 25e36. https://doi.org/10.1016/j.quascirev.2017.03.017. Marshall, F., 1989. Rethinking the role of Bos indicus in Sub-Saharan Africa. Current Scroxton, N., Burns, S.J., Dawson, P., Rhodes, J.M., Brent, K., McGee, D., Heijnis, H., Anthropology 30(2), 235e240. Gadd, P., Hantoro, W., Gagan, M.K., 2018. Rapid measurement of strontium in Matsumoto, K., Burney, D.A., 1994. Late Holocene environments at Lake Mitsinjo, speleothems using core-scanning micro X-ray fluorescence. Chemical Geology northwestern Madagascar. The Holocene 4, 16e24. 487, 12e22. McDermott, F., 2004. Palaeo-climate reconstruction from stable isotope variations Trouet, V., Van Oldenborgh, G.J., 2013. KNMI Climate Explorer: A Web-Based in speleothems: a review. Quaternary Science Reviews 23, 901e918. Research Tool for High-Resolution Paleoclimatology, 69, pp. 3e13. https:// Meador, L.R., Godfrey, L.R., Rakotondramavo, J.C., Ranivoharimanana, L., Zamora, A., doi.org/10.3959/1536-1098-69.1.3. Sutherland, M.R., Irwin, M.T., 2017. Cryptoprocta spelea (Carnivora: Eupleridae): Verschuren, D., Laird, K.R., Cumming, B.F., 2000. Rainfall and drought in equatorial What did it eat and how do we know? Journal of Mammalian Evolution. https:// east Africa during the past 1100 years. Nature 403, 410e414. https://doi.org/ doi.org.silk.library.umass.edu/10.1007/s10914-017-9391-z. 10.1038/35000179. Muldoon, K.M., Crowley, B.E., Godfrey, L.R., Rasoamiaramanana, A., Aronson, A., Virah-Sawmy, M., Willis, K.J., Gillson, L., 2009. Threshold response of Madagascar's Jernvall, J., Wright, P.C., Simons, E.L., 2012. Early Holocene fauna from a new littoral forest to sea-level rise. Global Ecology and Biogeography 18, 98e110. subfossil site: A first assessment from Christmas River, south central Virah-Sawmy, M., Willis, K.J., Gillson, L., 2010. Evidence for drought and forest Madagascar. Madagascar Conservation & Development 7, 23e29. declines during the recent megafaunal extinctions in Madagascar. Journal of Parker-Pearson, M., Godden, K., Ramilisonina, Retsihisatse, Schwenninger, J.L., Biogeography 37, 506e519. Heurtebize, G., Radimilahy, C., Smith, H., 2010. Pastoralists, Warriors and Voarintsoa, N.R.G., Wang, L., Railsback, L.B., Brook, G.A., Liang, F., Cheng, H., Colonists: The Archaeology of Southern Madagascar. Archaeo Press, Oxford. Edwards, R.L., 2017. Multiple proxy analyses of a U/Th-dated stalagmite to Perez, V.R., Burney, D.A., Godfrey, L.R., Nowak-Kemp, M., 2003. Butchered sloth reconstruct paleoenvironmental changes in northwestern Madagascar between lemurs. In: Godfrey, L.R., Jungers, W.L. (Eds.), The extinct Sloth Lemurs of 370CE and 1300CE. Palaeogeography Palaeoclimatology Palaeoecology 469, Madagascar. Evolutionary Anthropology, vol. 12, p. 260. 138e155. Perez, V.R., Godfrey, L.R., Nowak-Kemp, M., Burney, D.A., Ratsimbazafy, J., Vasey, N., Wang, L., Brook, G.A., 2013. Holocene climate changes in northwest Madagascar: evi- 2005. Evidence of early butchery of giant lemurs in Madagascar. Journal of dence from a two-meter-long stalagmite from the Anjohibe Cave. In: 2013 Meeting Human Evolution 49, 722e742. Program of the Association of American Geographers published online, session 1512. Pierron, D., Heiske, M., Razafindrazaka, H., Rakotob, I., Rabetokotany, N., Winterhalder, B., Goland, C., 1993. On population, foraging efficiency, and plant Ravololomang, B., Rakotozafy, L.M.-A., Rakotomalala, M.M., Razafiarivony, M., domestication. Current Anthropology 34, 710e715. Rasoarifetra, B., Andriamampianina Raharijesy, M., Razafindralambo, L., Ram- Wong, C.I., Breecker, D.O., 2015. Advancements in the use of speleothems as climate ilisonina, Fanony, F., Lejamble, S., Thomas, O., Abdallah, A.M., Rocher, C., archives. Quaternary Science Reviews 127, 1e18. https://doi.org/10.1016/ Arachiche, A., Tonaso, L., Pereda-loth, V., Schiavinato, S., Brucato, N., Ricaut, F.- j.quascirev.2015.07.019. X., Kusuma, P., Sudoyo, H., Ni, S., Boland, A., Deleuze, J.-F., Beaujard, P., Zahid, H.J., Robinson, E., Kelly, R.L., 2016. Agriculture, population growth, and sta- Grange, P., Adelaar, S., Stoneking, M., Rakotoarisoa, J.A., Radimilahy, C., tistical analysis of the radiocarbon record. Proceedings of the National Academy Letellier, T., 2017. Genomic landscape of human diversity across Madagascar. of Sciences of United States of America 113, 931e935. Proceedings of the National Academy of Sciences of United States of America Zinke, J., Pfeiffer, M., Timm, O., Dullo, W.C., Davies, G.R., 2005. Atmosphere-ocean 114(32), E6498eE6506. https://doi.org/10.1073/pnas.1704906114. dynamics in the Western Indian Ocean recorded in corals. Philosophical Pomerantz, S.M., 2017. The prehistory of Madagascar: Microbotanical and archae- Transactions of the Royal Society of London A: Mathematical, Physical and ological evidence from coastal and highland Sites. PhD dissertation, The Uni- Engineering Sciences 363, 121e142. https://doi.org/10.1098/rsta.2004.1482. versity of Oxford. Zinke, J., Loveday, B.R., Reason, C.J.C., Dullo, W.C., Kroon, D., 2014. Madagascar corals Prendergast, M.E., Buckley, M., Crowther, A., Frantz, L., Eager, H., Lebrasseur, O., track sea surface temperature variability in the Agulhas Current core region et al., 2017. Reconstructing Asian faunal introductions to eastern Africa from over the past 334 years. Scientific Reports 4, 4393. https://doi.org/10.1038/ multi-proxy biomolecular and archaeological datasets. PLoS One 12(8), srep04393. e0182565.